
All the output directories were tar'ed and compressed with zstd to reduce file
size and number. Each directory contains the Alphafold2 2.2.0 output for one of
the OPGs. This includes 
- MSAs
- pickle files for input features
- 5 predicted unrelaxed structures (PDB),
- a relaxed structure for each of the 5 predictions (PDB)
- the relaxed structures renamed after ordering them by the mean pLDDT
  of the structure
- 5 pickle files for the results of the prediction (e.g. per residue pLDDT,
  ...)

The archive files were created with

export ZSTD_NBTHREADS=6
export ZSTD_CLEVEL=19
for d in $(find . -maxdepth 1 -mindepth 1 -type d) ; do
    tar -I zstd -cf "${d}.tar.zst" "$d"
done

All but 2 files were 2GB or less. 2 (opg194 and opg210) were 4GB. For these two
the archives were redone after removing the result pickle files for all but the
top scoring prediction.

Directories can be reconstituted with

for f in *.tar.zst ; do
    tar -I zstd -xf $f
done


Compressed data size: 40GB
Uncompressed data: 68G 
